地球信息科学学报 ›› 2011, Vol. 13 ›› Issue (3): 305-312.doi: 10.3724/SP.J.1047.2011.00305

• 土地覆被与土地利用研究 • 上一篇    下一篇

干旱荒漠区植被覆盖变化的遥感监测分析

崔耀平1,2, 刘彤1, 赵志平2, 李佳3   

  1. 1. 石河子大学生命科学学院,石河子 832000;
    2. 中国科学院地理科学与资源研究所,北京 100101;
    3. 中国科学院新疆生态与地理研究所,乌鲁木齐 830011
  • 收稿日期:2011-02-22 修回日期:2011-05-30 出版日期:2011-06-25 发布日期:2011-06-15
  • 通讯作者: 刘 彤(1968-),男,博士,教授。主要从事植被生态和进化生物学研究。E-mail:liutong1968@yahoo.com.cn E-mail:liutong1968@yahoo.com.cn
  • 作者简介:崔耀平(1984-),男,博士研究生,主要从事遥感应用与环境信息模拟研究。E-mail:cuiyp@lreis.ac.cn
  • 基金资助:

    国家十一五科技支撑计划重大项目(2007BAC17B03)资助。

Using Multi-spectral Remote Sensing Data to Extract and Analyze the Vegetation Change of the Western Gurbantunggut Desert

CUI Yaoping1,2, LIU Tong1, ZHAO Zhiping2, LI Jia3   

  1. 1. College of Life Sciences, Shihezi University, Shihezi 832000, China;
    2. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    3. Xinjiang Institute of Ecology and Geography, CAS, Urumqi 830011, China
  • Received:2011-02-22 Revised:2011-05-30 Online:2011-06-25 Published:2011-06-15

摘要: 多 期的遥感数据可以用来分析干旱荒漠区植被的空间分布格局和变化特征。本文以1989、2000和2007年3个不同时相的Landsat TM/ETM+ 影像为数据源,利用线性光谱混合分析和RGB彩色合成法构建一个研究框架,对古尔班通古特沙漠西缘进行植被信息的提取和变化监测分析。在混合像元分解过程中,通过多种方法选择端元,比较选出最佳的端元数目及其对应光谱特征值,对植被变化监测的结果,结合气象等因子综合判定。结果表明:(1)研究区内的植被、盐碱地、裸沙和黑色砂砾等4种端元被选取出来,分析表明非受限的最小包含端元特征法所选端元光谱特征的分解结果较为理想;(2)以线性光谱混合分解技术提取的干旱荒漠区植被分量与实测植被盖度显著相关,线性相关系数为0.86,可见干旱荒漠区的植被盖度可以通过遥感影像提取的植被分量间接得到;(3)研究时段内,研究区植被覆盖变好区域占研究区总面积的41.47%,而退化区域仅占16.51%,综合分析结果也说明植 被总体情况呈现好转。

关键词: 干旱荒漠区, 植被变化, 古尔班通古特沙漠, 端元, 光谱混合分析

Abstract: Acquiring vegetation information is a key step in monitoring and evaluating arid land cover, even though it is difficult to extract desert vegetation information due to the confounding effort of soil or sand. There are many limiting factors for estimating vegetation growth in arid landscapes. Nevertheless, the growth of perennial vegetation reflects comprehensive vegetation conditions and can represent vegetation cover during a certain period. So in this study, we analyzed three years of Landsat images (1989, 2000 and 2007) that covered a typical portion of the Gurbantunggut Desert to estimate the vegetation change that has occurred. After comparing different methods, we finally chose the U-min method (one of the spectral mixture analysis methods) as the best way to determine the fraction of vegetation information of Landsat images. The RGB composition method was used to monitor the changes in vegetation abundance. In the end, we also comprehensively estimated vegetation changes using annual average time-series precipitation data and other's NDVI research results. The main results showed that: (1) a significant linear relationship exists between vegetation cover and vegetation fraction, with a correlation coefficient of 0.858. So the vegetation cover percent can be expressed by vegetation fraction extracted from remote sense images. (2) The area with improved vegetation cover accounted for 41.47% of the whole study area, while these patches with degraded vegetation cover accounting for 16.51%. However, it should be noted that these degraded vegetation pixels were distributed more sporadically. (3) During the study period, improved vegetation regions were mainly located in the semi-fixed dunes. At the same time, the analysis results also showed that the vegetation was distributed extensively in the arid land, regions including vegetation grown occupied almost 90% of the whole study area, and the number of improved vegetation pixels was greater than the number of degraded vegetation pixels during the period of 20 years.

Key words: endmember, spectral mixture analysis, arid area, vegetation change, Gurbantunggut Desert